Markopolo

Athena is Markopolo's in-house model, described by the company as a 4B parameter Large Event Model. It reads sequences of shopping events, such as repeated size guide views or price page revisits, and predicts the next action along with the reason behind it. The practical output is a purchase probability score that decides whether outreach fires now or waits. Markopolo says the model topped the OPeRA benchmark, a claim that hasn't been independently verified.

Three recovery plays ship by default: browse, cart, and checkout abandonment. Each one runs without a hand-built workflow or a line of A/B tests. When a shopper stalls, the agent picks the channel and the message, then follows up if the first attempt gets no response. That escalation path is the part most stores can't staff manually, since it needs someone watching the clock.

Email, SMS, WhatsApp marketing, push notifications, and AI voice calls all run from the same dashboard. Channel choice is part of the model's decision, not a setting a marketer locks in advance. AI voice minutes are capped by plan: 10 minutes on Standard and 60 on Pro. A shopper who ignores email might get a WhatsApp note and then a call, which is close to how a salesperson would work a lead.

MarkTag is a composable customer data platform that ships with the product. It stitches anonymous sessions to known profiles and keeps context when a customer moves between channels. Stores that already run a CDP can connect it, while smaller teams can skip the extra purchase. Unified profiles are the prerequisite for 1:1 personalization, so this layer matters more than it sounds.

Discounts get expensive when they go to shoppers who would have bought anyway. The predictive engine scores price sensitivity and offer responsiveness, then reserves incentives for people who need a push. Markopolo shows predicted lift per shopper in its interface. For a store with thin margins, this is the feature that pays for the subscription.

Dashboards track conversion rate, average order value, engagement, email and SMS click rates, and attributed revenue. Attribution is the number finance teams ask about first, and having it next to campaign performance saves a step of exporting and reconciling figures. Reporting depth is thinner than dedicated analytics tools, so heavy analysts will still pull data elsewhere.

The Shopify app installs from the app store and syncs catalog and order data. CRM connections bring historical customer records into the model so new campaigns start with context instead of a blank profile. Markopolo also offers free migration of customer data, audiences, automations, and campaigns for teams switching from another platform.